cucker tarlsonclocks and latency.
ryzen is taking mainstream into a totally different direction so no surprises.

Also why I returned my 1700 and went with 5Ghz 8700k for chart topping gaming performance that’ll actually last me for years instead of having to constantly upgrade with Ryzen Zen -> Zen+ -> Zen2 and STILL be slower than a 5 year old Intel arch lol. The CCX rumors better be true for AMD’s sakes.

laszlo

That's what happens when you try running your AMD 3000 series CPUs @INTEL @5GHz+ speeds lol

Base clock should have been 5 Ghz a long time ago. That being said lets see the benchmark comparisons.

Rob94hawkBase clock should have been 5 Ghz a long time ago. That being said lets see the benchmark comparisons.

Base clocks are still under 4GHz and only Turbo goes to 5GHz+

I'm running my 9900KS @5.2GHz AVX2 all cores so technically it's 5GHz base with 5.2GHz Turbo.

I'm betting 10700K is only running 2 cores 5.3GHz on Turbo OC. The CPUs are made for your average 200w CPU cooler.

GoldenXAMD is going to increase prices, get ready.

Highly unlikely, while they are selling as many as they can make right now the step up from 2990WX to 3990X (in terms of price) as well as 2700x to 3950x has allowed Intel to offer better VFM across lots of price points. Which basically means that if AMD do go even higher, especially for their mainstream flagship, their sales could go down at least in proportion to what they're currently enjoying in the DIT segment.

efikkanGames will not be able to scale well with many threads. The workloads of a game is basically divided into three categories; 1) rendering, 2) core game simulation 3) optional stuff (like networking, sound effects etc.). Rendering(1) only scales until the GPU is no longer bottlenecked. While some games can leverage multiple queues to do multi-pass rendering etc., the scaling potential here is fairly limited anyway, and the thread(s) feeding the GPU is usually the only ones which matters for FPS anyway. Also, the future long-term trend in games is that the GPU does more and more of the heavy lifting. Game simulation(2)(game loop) is usually a fixed workload, since the game needs to work the same across machines, and is usually scaled towards the minimum requirements of a game. Games are also incredible latency sensitive, which makes it very hard to divide timing critical tasks into tiny chunks across many cores. We might see some more multi-core scaling in gaming, but games will not use 16 threads even 10 years from no (except for a couple of edge cases, of course).

Applications in general do slowly scale better with multiple threads, but even here there are theoretical limits to what is achievable. Even with the best efforts, most synchronized workloads (like most client applications are) will see a scaling drop-off around 8 threads, and not scale well beyond 16 threads. Asynchronous workloads scale nearly perfectly though, but very few of those are relevant for end-users. Anything beyond 8 cores is more relevant for users running multiple applications simultaneously rather than single applications.

The CPUs in the new consoles arriving later in 2020 will have 8 cores and 16 threads.
Games can use as many threads as you throw at them, because there are areas like physics, AI acceleration, ray-tracing acceleration all of which will greatly benefit from many cores.

I can't wait to try the 64-core Threadripper 3990X or its derivatives shrunk to lower TDP envelopes.

Not that I would care, yet I had no clue if Intel was competing against nuclear fision , regarding thermal power density. Good for them , good for Intel.

ARFThe CPUs in the new consoles arriving later in 2020 will have 8 cores and 16 threads.
Games can use as many threads as you throw at them, because there are areas like physics, AI acceleration, ray-tracing acceleration all of which will greatly benefit from many cores.

I can't wait to try the 64-core Threadripper 3990X or its derivatives shrunk to lower TDP envelopes.

Gaming does not benefit from more cores yet!!! Thus the 9900KS is the fastest Gaming CPU! (And yes the 9900KS is the gaming king!) And yes the 9900KS will beat the 3990X in PC gaming. Plus the 3990X is 280w Watts monster if your talking about TDP and the mortgage you need to take out to buy one lol.

We are still living in a world of 4 cores still where the money making entry level CPUs thrive your i5 & R5 series people.

ToxicTaZGaming does not benefit from more cores yet!!! Thus the 9900KS is the fastest Gaming CPU! (And yes the 9900KS is the gaming king!) And yes the 9900KS will beat the 3990X in PC gaming. Plus the 3990X is 280w Watts monster if your talking about TDP and the mortgage you need to take out to buy one lol.

We are still living in a world of 4 cores still where the money making entry level CPUs thrive your i5 & R5 series people.

Why would compare a 64 core CPU that WIndows 10 sees as 2 CPUs with a CPU that is for gaming. You should reference the 3800X instead and is the 9900KS that much faster period..............?

ToxicTaZGaming does not benefit from more cores yet!!! Thus the 9900KS is the fastest Gaming CPU! (And yes the 9900KS is the gaming king!) And yes the 9900KS will beat the 3990X in PC gaming. Plus the 3990X is 280w Watts monster if your talking about TDP and the mortgage you need to take out to buy one lol.

We are still living in a world of 4 cores still where the money making entry level CPUs thrive your i5 & R5 series people.

Core i9-9900KS is fastest only if you game at 720p and 1080p with a RTX 2080 Ti.
If you game at 2160p with Radeon RX 5700 XT, there will be no difference.

So, it's better for you to buy the Ryzen 9 3900X that has 24 threads for everything beyond only gaming!


And yes, you have to run the 64-core Threadripper 3990X either under Windows 10 Enterprise or Linux.

ARFCore i9-9900KS is fastest only if you game at 720p and 1080p with a RTX 2080 Ti.
If you game at 2160p with Radeon RX 5700 XT, there will be no difference.

It's still ~10% faster than a mid-range Ryzen even at QHD: www.techpowerup.com/review/intel-core-i9-9900ks/15.html
Granted, even a flat 10% isn't much of a difference. I mean, if Ryzen is unplayable at 30fps in a game you like, 9900KS' 33fps won't help you much.

ToxicTaZGaming does not benefit from more cores yet!!! Thus the 9900KS is the fastest Gaming CPU! (And yes the 9900KS is the gaming king!) And yes the 9900KS will beat the 3990X in PC gaming. Plus the 3990X is 280w Watts monster if your talking about TDP and the mortgage you need to take out to buy one lol.

We are still living in a world of 4 cores still where the money making entry level CPUs thrive your i5 & R5 series people.

Why are you comparing a 64 core workstation CPU to an 8 core Mainstream desktop CPU? Per-Core power draw on the 3990X is 1/3rd what the 9900KS pulls:

ARFI think it will be an epic fail and the only way to escape from that punishment is not to buy it. Ever.

Especially when you can buy the 65-watt Ryzen 7 3700X for $310.



If this is true, the 10-core i9-10900K flagship will cost $800-$900. :laugh:

Who is pay $300+ for a 3700x? Microcenter had them last week for $259

GlacierNineWhy are you comparing a 64 core workstation CPU to an 8 core Mainstream desktop CPU? Per-Core power draw on the 3990X is 1/3rd what the 9900KS pulls:

In part the reason for this is that the best chiplets with lowest voltages go for the 3990X.

Imagine if the best chiplets were used for the lower end models. Can be possible a 45-55-watt 3700X.

DeeJay1001Who is pay $300+ for a 3700x? Microcenter had them last week for $259

Probably the tiny fraction of the world that doesn't live next to a Microcenter?

ARFIn part the reason for this is that the best chiplets with lowest voltages go for the 3990X.

Imagine if the best chiplets were used for the lower end models. Can be possible a 45-55-watt 3700X.

Sure, but I'm mostly just trying to get Tax to engage brain for once instead of being... well, the way he is.

He's basically trying to compare a Mazda 787B to something designed for the Paris-Dakar. Sure, the Mazda will dominate around a racetrack like say, Suzuka, but the Dakar buggy will race in circumstances where the 787B won't even succeed in leaving the starting line. Pretending they're built for the same job is beyond stupid, it displays a total inability to actually think.

GlacierNineWhy are you comparing a 64 core workstation CPU to an 8 core Mainstream desktop CPU? Per-Core power draw on the 3990X is 1/3rd what the 9900KS pulls:

There is quite a lot wrong with just putting out that graph.
1. This is based off the whole system consumption. 112W/64 looks much better than 59W/8 especially is almost all of it has nothing to do with CPU.
2. 3.3 GHz vs 4.4 GHz. Zen2 does 3.4-3.5GHz at 2-3W which is an awesome result. At the same time it does 4.2GHz at about 10W. Any higher than that it goes all out of whack. Mine does 4.3 GHz at about 15W.
This is core only consumption.

londisteThere is quite a lot wrong with just putting out that graph.
1. This is based off the whole system consumption. 112W/64 looks much better than 59W/8 especially is almost all of it has nothing to do with CPU.
2. 3.3 GHz vs 4.4 GHz. Zen2 does 3.4-3.5GHz at 2-3W which is an awesome result. At the same time it does 4.2GHz at about 10W. Any higher than that it goes all out of whack. Mine does 4.3 GHz at about 15W.
This is core only consumption.

Sure, but is there any more wrong with that graph than there is with trying to claim that a 3990X and a 9900KS are competing products? Because that graph is more than good enough to prove the point I was making.

MistralCan anyone please confirm if this comes boxed with a chiller or not?..

K-models don't, thankfully. ;)

Rob94hawkBase clock should have been 5 Ghz a long time ago. That being said lets see the benchmark comparisons.

And Netburst(Pentium 4) was supposed to reach 7-8 GHz…
The long-term trajectory for upcoming nodes is decreased clock speed, and all of these CPUs are pushing clocks into "throttle territory", so pushing clocks significantly higher is not feasable until different types of semi-conductors arrive on the market.

The way forward is higher IPC and more SIMD.

ARFThe CPUs in the new consoles arriving later in 2020 will have 8 cores and 16 threads.
Games can use as many threads as you throw at them, because there are areas like physics, AI acceleration, ray-tracing acceleration all of which will greatly benefit from many cores.

I think you missed the point. It's not about just spawning threads, but to actually use them for performance gains. Interestingly, all the things you mentioned could or should be accelerated on a GPU, not on a bunch of CPU cores.

The key to multithreaded scaling is dividing a workload into independent work chunks and let the cores chew at them, this is very difficult for games, as they are a pipeline of serial tasks, where only some smaller bits can be parallelized, but all of them needs to be synced up many times throughout the pipeline. For a game with a tick rate of e.g. 120 Hz, there is an 8.33ms window for the entire game simulation, usually starting with input event processing followed by game logic like collisions etc. If the game uses particle simulations etc. for effects, this has to come after the core game simulation, but before the rendering. The overhead of synchronizing threads grows with thread count, so there will be a point where you get into diminishing returns and not to mention stutter.

And when it comes to the rendering, there is really no point in throwing more threads at it, as the CPU only needs to keep the GPU busy. While it is possible to have multiple CPU threads build a single queue, there is no point to it as it will only create more latency on the driver side, not to mention synchronization problems. The only real purpose with multiple threads for rendering is to do different tasks, like having one thread to load resources while another is rendering. There will be very few real-world cases where more than 2-3 threads would be useful for the GPU in gaming.

efikkanI think you missed the point. It's not about just spawning threads, but to actually use them for performance gains. Interestingly, all the things you mentioned could or should be accelerated on a GPU, not on a bunch of CPU cores.

The key to multithreaded scaling is dividing a workload into independent work chunks and let the cores chew at them, this is very difficult for games, as they are a pipeline of serial tasks, where only some smaller bits can be parallelized, but all of them needs to be synced up many times throughout the pipeline. For a game with a tick rate of e.g. 120 Hz, there is an 8.33ms window for the entire game simulation, usually starting with input event processing followed by game logic like collisions etc. If the game uses particle simulations etc. for effects, this has to come after the core game simulation, but before the rendering. The overhead of synchronizing threads grows with thread count, so there will be a point where you get into diminishing returns and not to mention stutter.

And when it comes to the rendering, there is really no point in throwing more threads at it, as the CPU only needs to keep the GPU busy. While it is possible to have multiple CPU threads build a single queue, there is no point to it as it will only create more latency on the driver side, not to mention synchronization problems. The only real purpose with multiple threads for rendering is to do different tasks, like having one thread to load resources while another is rendering. There will be very few real-world cases where more than 2-3 threads would be useful for the GPU in gaming.

You must use the CPU cores which are mostly sitting idle during gaming. In some type of crossfired-hybrid acceleration CPU-GPU.
What you want is that the GPU takes the whole load, while the CPU is completely offloaded.

You have 64-core 3990X and better start coding for it now.

At the end of the day.... The 10700K should sit somewhere in between 3800X and 9900KS performance wise.

Intel will keep the fastest Gaming CPU title until the 10900KS or AMD comes out with a 5GHz Ryzen 4000.....it's GHz that wins in gaming not cores for now.

You're market cores are based upon your entry level CPUs and both AMD and Intel are still making 4 & 6 cores CPUs. Telling everyone they need more than 8 cores CPU is Ludacris at the moment.... Until your bottom end i5 & R5 get 8 cores standard.

This thread is Intel dual channel based should keep with in dual channel performance topics.

ToxicTaZAt the end of the day.... The 10700K should sit somewhere in between 3800X and 9900KS performance wise.

Intel will keep the fastest Gaming CPU title until the 10900KS or AMD comes out with a 5GHz Ryzen 4000.....it's GHz that wins in gaming not cores for now.

You're market cores are based upon your entry level CPUs and both AMD and Intel are still making 4 & 6 cores CPUs. Telling everyone they need more than 8 cores CPU is Ludacris at the moment.... Until your bottom end i5 & R5 get 8 cores standard.

This thread is Intel dual channel based should keep with in dual channel performance topics.

There is absolutely no way in hell you are not an intentional troll.

ToxicTaZAt the end of the day.... The 10700K should sit somewhere in between 3800X and 9900KS performance wise.

Intel will keep the fastest Gaming CPU title until the 10900KS or AMD comes out with a 5GHz Ryzen 4000.....it's GHz that wins in gaming not cores for now.

You're market cores are based upon your entry level CPUs and both AMD and Intel are still making 4 & 6 cores CPUs. Telling everyone they need more than 8 cores CPU is Ludacris at the moment.... Until your bottom end i5 & R5 get 8 cores standard.

This thread is Intel dual channel based should keep with in dual channel performance topics.

Hmm I thought AMD got rid of 4 cores (APUs excluded) by 2nd gen there were no quad core the lowest was the 2600 and they are all hyper threaded. Not all games are based on clock speed and some games do gain from having multi cores (AOTS, Strange Brigade). The current thought process on gaming is 6 cores are the sweet spot. As a result no youtubers advise their watchers to use the I5 series.

kapone32Not all games are based on clock speed and some games do gain from having multi cores (AOTS, Strange Brigade). The current thought process on gaming is 6 cores are the sweet spot.

There is no such thing as optimizing for clock speed. Any scalable code will scale with core performance, no matter how that performance is achieved. If clock speed was the thing that mattered, Bulldozer would be the king of gaming CPUs!

And as many people keep forgetting; per core performance and multi-core performance are not counterparts, in fact faster cores are important to scale further with multithreading, as there are always diminishing returns.

Yawn, Intel.